1 /*
2 * Copyright (c) 2017-2020 Arm Limited.
3 *
4 * SPDX-License-Identifier: MIT
5 *
6 * Permission is hereby granted, free of charge, to any person obtaining a copy
7 * of this software and associated documentation files (the "Software"), to
8 * deal in the Software without restriction, including without limitation the
9 * rights to use, copy, modify, merge, publish, distribute, sublicense, and/or
10 * sell copies of the Software, and to permit persons to whom the Software is
11 * furnished to do so, subject to the following conditions:
12 *
13 * The above copyright notice and this permission notice shall be included in all
14 * copies or substantial portions of the Software.
15 *
16 * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
17 * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
18 * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
19 * AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
20 * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
21 * OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
22 * SOFTWARE.
23 */
24 #include "arm_compute/core/GLES_COMPUTE/kernels/GCPixelWiseMultiplicationKernel.h"
25
26 #include "arm_compute/core/Error.h"
27 #include "arm_compute/core/GLES_COMPUTE/GCHelpers.h"
28 #include "arm_compute/core/GLES_COMPUTE/GCKernelLibrary.h"
29 #include "arm_compute/core/GLES_COMPUTE/IGCTensor.h"
30 #include "arm_compute/core/GLES_COMPUTE/OpenGLES.h"
31 #include "arm_compute/core/Helpers.h"
32 #include "arm_compute/core/TensorInfo.h"
33 #include "arm_compute/core/Validate.h"
34 #include "arm_compute/core/Window.h"
35 #include "src/core/helpers/AutoConfiguration.h"
36 #include "src/core/helpers/WindowHelpers.h"
37 #include "support/StringSupport.h"
38
39 #include <cmath>
40 #include <cstdlib>
41 #include <set>
42 #include <string>
43 using namespace arm_compute;
44
GCPixelWiseMultiplicationKernel()45 GCPixelWiseMultiplicationKernel::GCPixelWiseMultiplicationKernel()
46 : _input1(nullptr), _input2(nullptr), _output(nullptr)
47 {
48 }
49
configure(const IGCTensor * input1,const IGCTensor * input2,IGCTensor * output,float scale)50 void GCPixelWiseMultiplicationKernel::configure(const IGCTensor *input1, const IGCTensor *input2, IGCTensor *output, float scale)
51 {
52 ARM_COMPUTE_ERROR_ON_NULLPTR(input1, input2, output);
53 ARM_COMPUTE_ERROR_ON_DATA_TYPE_CHANNEL_NOT_IN(input1, 1, DataType::F32);
54 ARM_COMPUTE_ERROR_ON_MISMATCHING_DATA_TYPES(input1, input2);
55 ARM_COMPUTE_ERROR_ON_NULLPTR(output);
56 ARM_COMPUTE_ERROR_ON_MSG(scale < 0, "Scale cannot be negative. ");
57
58 // Auto initialize output if not initialized
59 {
60 set_shape_if_empty(*output->info(), input1->info()->tensor_shape());
61 set_format_if_unknown(*output->info(), Format::F32);
62 }
63
64 ARM_COMPUTE_ERROR_ON_MISMATCHING_SHAPES(input1, input2, output);
65 ARM_COMPUTE_ERROR_ON_MISMATCHING_DATA_TYPES(input1, input2, output);
66 ARM_COMPUTE_ERROR_ON_MSG(scale < 0, "Scale cannot be negative. ");
67
68 _input1 = input1;
69 _input2 = input2;
70 _output = output;
71
72 std::string data_type;
73 std::string compute_type;
74
75 // Set kernel build options
76 std::set<std::string> build_opts;
77 build_opts.emplace("#define LOCAL_SIZE_X " + support::cpp11::to_string(1));
78 build_opts.emplace("#define LOCAL_SIZE_Y " + support::cpp11::to_string(1));
79 build_opts.emplace("#define LOCAL_SIZE_Z " + support::cpp11::to_string(1));
80
81 build_opts.emplace("#define SCALE " + support::cpp11::to_string(scale));
82
83 // Create kernel
84 _kernel = static_cast<GCKernel>(GCKernelLibrary::get().create_kernel("pixelwise_mul_float", build_opts));
85
86 // Configure kernel window
87 constexpr unsigned int num_elems_processed_per_iteration = 1;
88
89 Window win = calculate_max_window(*input1->info(), Steps(num_elems_processed_per_iteration));
90
91 AccessWindowHorizontal input1_access(input1->info(), 0, num_elems_processed_per_iteration);
92 AccessWindowHorizontal input2_access(input2->info(), 0, num_elems_processed_per_iteration);
93 AccessWindowHorizontal output_access(output->info(), 0, num_elems_processed_per_iteration);
94
95 update_window_and_padding(win, input1_access, input2_access, output_access);
96
97 ValidRegion valid_region = intersect_valid_regions(input1->info()->valid_region(),
98 input2->info()->valid_region());
99 output_access.set_valid_region(win, valid_region);
100
101 IGCKernel::configure(win);
102 }
103
run(const Window & window)104 void GCPixelWiseMultiplicationKernel::run(const Window &window)
105 {
106 ARM_COMPUTE_ERROR_ON_UNCONFIGURED_KERNEL(this);
107 ARM_COMPUTE_ERROR_ON_INVALID_SUBWINDOW(IGCKernel::window(), window);
108
109 _kernel.use();
110
111 Window slice = window.first_slice_window_3D();
112
113 do
114 {
115 unsigned int idx = 0;
116 unsigned int binding = 1;
117 add_3D_tensor_argument(idx, _input1, binding++, slice);
118 add_3D_tensor_argument(idx, _input2, binding++, slice);
119 add_3D_tensor_argument(idx, _output, binding++, slice);
120
121 _kernel.update_shader_params();
122 enqueue(*this, slice);
123 }
124 while(window.slide_window_slice_3D(slice));
125 }
126